SSSWG - March 1995 Minutes
IEEE Storage System Standards Working Group Meeting
14-17th March, 1995
Digital Equipment Corporation, Colorado Springs, Colorado.
Attendance:
Carl Braganza
Jack Cole
Anat Gafni
Bruce K. Haddon
Creig Humes
Merritt Jones
Dennis Luck
John Merrill
Gary Mueller
Michael T. Peterson
Leo Profilet
Rich Ruef
Dave Skinner
Eric Stouffler
Hamid Vazire
Lee Wand
Bill Watson
Joel Williams
Rich Wrenn
Chair: Dave Skinner.
Vice Chair: Jack Cole.
Recording Secretary: Bruce K. Haddon.
General Session, Tuesday 14 March, 9.00 a.m.
Agenda
======
General Session
Announcements
Minutes of the previous meeting
Treasurer's report
Review of Action Items from previous meeting
Status of recruitment and sponsorship.
Status of RFP's
Discussion of Proposal for a Version 6 OSSI Model
Presentation and discussion of the PVL Security Model (Michael
Peterson)
Presentation and discussion on API style (Rich Wrenn)
Presentation and discussion of overview of VSS (Bob Baird)
Presentation of EOSDIS project (Joel Williams)
Subcommittee Sessions
Closing General Session
Subcommittee status (each Chairman)
Tabled Items
Review of Open Action Items
Proposed Agenda for 95/5 Meeting
Other business
1. Announcements
=============
* Welcome to new attendees:
Leo Profilet
Eric Stouffler
Lee Wand
* UNIFORUM occurs next week.
* Subcommittee meetings will be:
MVR
PVL
PVR
* Fourth NASA Goddard Conference on Mass Storage Systems and
Technologies will be held at the University of Maryland, College
Park, Maryland.
* Network Storage `95: A Conference on Building High Data
Accessibility Networks, June 5-7, 1995, Santa Clara. Contact
Strategic Research Corporation, 350 So. Hope Avenue, A-103,
Santa Barbara, Ca 93105, (805) 569-5610.
2. Minutes of previous meeting
===========================
## Resolved: that the minutes of the previous meeting be approved
with the following corrections:
a) Jack Cole's remark with respect to "MIC" was not that
we "compete" but work "in parallel to" their
activities.
b) The reference to "end of life" of SSSWG reflects an
intent to consider at some point in time planning for
"end of life", not any imminent event of this type.
Moved: Mike Peterson; 2nd: Jack Cole.
Carried unanimously.
* Joel Williams name was inadvertently omitted from the attendance
list for the previous meeting.
3. Treasurer's Report
==================
Dave Skinner reported on behalf of Thomas Jefferson. Checks from last
meeting should have cleared. There were little expenses from last
meeting. Cash position is, therefore, good. No attempt has been made to
collect the ANSI "tax", so we haven't paid.
4. Review of Action Items from Previous Meeting
============================================
1. First draft of an SSSWG Business Plan (overall
objectives for the three key dot groups, V6 progress,
etc.)
Responsible: Dave Skinner and Jack Cole
Due: 31 January
Status: Open
## Resolved: change "business" to "operational".
Moved: Dennis Luck, 2nd Mike Peterson
Carried without dissent.
2. Subcommittee Management Plan (draft standards, work in
progress, notes, minutes)
Responsible: Dave Skinner and Jack Cole
Due: 31 January
Status: Open.
3. Locate a replacement for "swedishchef".
Responsible: Dave Skinner
Due: (not specified)
Status: Open.
4. Report POSIX status, and make available copies of PAR's
on interacting areas.
Responsible: Dave Skinner
Due: By end of 95/01 meeting
Status: Completed
5. First draft of a process for "internal" balloting
within the SSSWG. This process is needed in order for
the SSSWG to reach consensus on the draft which is to be
submitted to IEEE balloting.
Responsible: Jack Cole
Due: 95/03 meeting
Status: Open
There were many minutes of discussion reflecting the
importance of this question. Jack agreed to continue
this item.
6. First draft of a process to address questions of
lifetime, size, logging, evaluation, etc. for RFP's.
Responsible: Dave Skinner and Jack Cole
Due: 27 January
Status: Open
5. Status of recruitment and sponsorship
=====================================
Possibly new membership from IBM. Possibility of representation from
Exabyte, and perhaps E-Mass (Denver).
Jack Cole asked about material for Goddard.
6. Status of RFP's
===============
* OID Issued: Bob Baird (VSS)
One response and subsequent discussion.
* "Plug and Play" Not issued.
* Base Types
Issued: Mike Peterson (PVL)
One response
* Criteria: consistency requirements between model components
Issued: Mike Peterson.
No responses.
* Naming conventions. Not issued.
* API calling conventions (argument passing) Not issued.
* Exception handling / Error propagation Not issued.
Discussion: Noted need to complete the definition of the process for
handling RFP's. Also noted: there are dependencies between RFP's.
## Resolved: a sequencing of action items needs to be established,
or dependencies determined, for the issuing of the RFP's.
Moved: Carl Braganza, 2nd Anat Gafni.
Discussion: Mike Peterson expressed concern that the question of
"action items" and "dependencies" are linked in the motion.
Motion was withdrawn by the mover and seconder.
## Resolved: that the "suggestions" for RFP's from the previous
meeting be turned into "action items".
Moved: Eric Stouffler; 2nd, Mike Peterson
Carried without dissent
# Action Item: RFP for "Plug and Play" to be issued.
David Skinner (PVR)
# Action Item: RFP for Naming conventions to be issued.
Craig Humes (MVR)
# Action Item: RFP for API calling conventions (argument passing)
to be issued.
David Skinner (PVR)
# Action Item: RFP for Exception handling / Error propagation to
be issued.
Craig Humes (PVR)
7. Discussion of Proposal for a Version 6 OSSI Model
=================================================
Discussion: various people made the following comments:
There is a need to record problems and inconsistencies found in the V5
model as the standards work proceeds. Already there has been
considerable work lost as parts of the V5 draft were deleted in order to
gain consensus for the "approved draft" designation.
The only way to gain consistency in a model document is to have a "Chief
architect" who is more than just an editor. This person must have
sufficient understanding, and time, to work the consistency issues with
each of the subcommittees.
The goals for a V6 model would be to act as a method of tracking changes
that occur as the standard API's are developed. However, we need to
develop credibility that standards will emerge: we must not divert
energy that would otherwise be devoted to developing the API's.
A suggestion was made that one-seventh of the meeting time, e.g., Friday
morning of a three-and-a-half day meeting, be used for V6 work.
## Resolved: that we defer all technical editing work on V6 model
till after the September MSS meeting
Moved: Dave Skinner; 2nd, Rich Wrenn
Discussion: it has been suggested that the technical editing
work be the recording of architectural changes, hence this
motion will interfere with that. Mike Peterson feels that the
changes are obvious from the API work. Joel asked whether a
"bulletin board" might not be a method of tracking the changes.
Carried, with one dissent.
## Resolved: that an action item be assigned to implement the
process to record changes to the V5 model.
Moved: Bill Watson; 2nd, Rich Ruef.
Discussion: Bruce Haddon pointed out the items cannot be
"assigned" in a volunteer group. Rich Wrenn felt that we are too
thin on resource to do this. Mike Peterson agreed, and restated
with his belief that the differences between the API's and the
model would be evident. Anat Gafni was not so sure that these
differences would be so obvious.
Motion failed 5-6.
The order of the following items were rearranged by the Chair to cater
to the needs of the presenters.
8. Presentation of EOSDIS project
==============================
Joel Williams presented an overview of the NASA Earth Observing System
Data and Information System (EOSDIS). Materials are available from him.
9. Presentation and Discussion of the PVL Security Model
=====================================================
Michael Peterson presented an overview of his investigations into
security issues as they relate to the PVL. Further details are available
in the paper submitted to the reflector last week.
Security here is primarily "reliability" (protecting the data against
failure in the media) and "integrity" (protection of the data against
accidental or deliberate corruption). The work being done in POSIX on
security models and access control lists (ACL's) is relevant: they are
into their 15th draft. It speaks only to (discretionary) access control,
not to the authentication problem (which SSSWG assumes is part of the
environment). In POSIX, ACL's are applied only to directories and files.
Discussion: Having a consistent security model across all 1244
components is needed, and, in particular, consistent so that higher
level components can implement their security in terms of the features
of the lower levels.
Dave Skinner asked whether it might be possible to define a security
model (for 1244) that did not have to reference a particular method of
implementation (e.g., ACL's).
The POSIX model maps all of the existing UNIX "permission" operations
into operations on ACL's.
10. Presentation and Discussion of API Style
========================================
Rich Wrenn asked to defer presentation in order to do a more detailed
presentation later in the meeting, but made the following introduction:
The style presented by Michael Peterson at the last meeting was "N
objects * M attributes" in terms of entry points, and required an
individual call per set or get of each attribute. He has a model that
avoids these draw backs.
11. Discussion of "Plug and Play"
=============================
By consent, Rich was asked to lead a discussion of the "plug and play"
issue, partly to refresh understanding, and to educate new members.
Discussion:
The first issue to receive attention was how should a PVL address more
than one PVR (although it was questioned whether this was a real
requirement). This is a simpler issue than full "plug and play" (PnP).
Defining an API alone does not solve the addressing issues involved in
having a PVL talk to more that one PVR. Carl suggested that this could
be resolved by defining "levels of compliance", whereby there would be,
say, an API, as well as an "on-the-wire" protocol.
So the essential question is how can multiple, lower-level entities be
addressed?
Some of the problem with defining the solution to this matter is that
the requirements change depending upon whether one is a customer, an
integrator, or a supplier. Historically, the makeup of the SSSWG has
moved from customer-domination to supplier-domination, to what is now
probably integrator-domination, as most suppliers now have strong
integration components.
Joel recommended doing a "reference" standard (with implementation)
based on DCE.
It has been long agreed that dependencies on "non-standards" (e.g., DCE,
X/Open, ONC RPC) should be specified within "profiles". But it is still
not clear that this resolves the problem.
David Skinner reminded everybody that the object stated in each of the
PAR's is that the 1244 subcommittees would define API's. The PnP and
related interoperability issues are "noble", but as they not stated
objectives in the PAR's, they should not be allowed to stop progress
(unless we are willing to undertake redefining the PAR's).
Rich Wrenn: if the API is defined assuming an RPC implementation, then
the "profile" (the documentation describing how to map the API into an
RPC implementation) becomes a small, straightforward task. Otherwise it
can be a very large and difficult job.
Session adjourned at 5:17 p.m., Tuesday, 14th March, 1995.
Wednesday, 15th March, 1995 was devoted to subcommittee meetings. The
PVL, PVR, and MVR subcommittees met during the day.
Session opened 9:03 a.m., Thursday, 16th March, 1995.
Agenda items for this session:
Expiry of PAR's: Merritt Jones.
API Style: Bruce K. Haddon
12. Expiry of PAR's
===============
Merritt stated that the PAR that will expire, dating from September,
1991, is titled "Guide for Storage System Design", and is numbered
simply "1244". This is the PAR for the Reference Model, otherwise known
as "dot 0".
There are six other PAR's: the "dot 1" through "dot 6" PAR's of 1244.
1244.1 OID
1244.2 PVL active
1244.3 PVR active
1244.4 MVR active
1244.5 MGT
1244.6 VSS active
Merritt asked that there be a clear commitment from the working group to
each of these PAR's, and that those suffering from "benign neglect" be
retired. Bruce Haddon felt that it was too soon to come to the
conclusion that there is "benign neglect". Joel Williams felt that
management(MGT) was really an issue with every other group. Dave Skinner
believes that as the other standards emerges, everybody's interest will
return to MGT. Mike Peterson stated his thought that the management PAR
was written too soon, before there were objects defined by the others to
manage.
On the other hand, the OID PAR appears to be too narrow and restricted
to be useful. The "OID" is just a single cross-model issue. Others are
API style, data types, etc. Rich Wrenn, joined by Mike Peterson, felt
the "dot 1" was not needed at all. Bruce Haddon and Dave Skinner felt
that having a place for these issues was useful, and Bruce in particular
would like to see "dot 1" PAR rewritten into a more general framework
standard.
There is a question about the lifetime of the "approved draft" of the
Reference Model, since a draft of a standard lives only a year. But the
Reference Model is a "guide", so the rules will have to be investigated.
# Action Item: Merritt will report to the next meeting on the
status of the "draft" of the Reference Model.
13. API Style
=========
Bruce Haddon led a presentation and discussion of the style factors
present in the PVR proposal made last year. Copies of the presentation
slides are available upon request.
Rich Wrenn then presented a code segment implementing a SHOW command to
demonstrated this style of API in use. The code segment was a complete
program for displaying attributes of physical volumes in a PVL or
cartridges in a PVR, in 5 1/2 pages.
There was considerable discussion during this presentation about style,
complexity, implementation methods, object orientation, code styles,
type checking, client/server architecture.
Rich Wrenn's closing remarks, agreed by Bruce, were that the SSSWG needs
very rapidly to converge on an API style. Mike Peterson pointed out that
there was really no agreement at this time that all model components
needed to use the same style, and, further, that we should not agree a
style that precludes the performance requirements of any of the
components.
Dave Skinner noted that the balloters on the "dots" would let us know
whether style consistency was required.
Session was adjourned for lunch at 12.10 PM, Thursday, 16 March, 1995.
Session resumed at 1.20 PM, Thursday, 16 March, 1995.
Session was adjourned at 1.22 PM, Thursday, 16 March, 1995.
The afternoon of Thursday, 16 March, was devoted to further meetings of
the subcommittees.
Session resumed at 9:00, Friday, 17 March, 1995.
14. IEEE P1244 SSSWG Operating Plan
===============================
Dave Skinner presented the outline for the proposed operating plan for
the 1244 SSSWG.
# Action Item: A copy of this outline will be posted to the
reflector.
Discussion: Joel Williams asked why we are undertaking for formulate an
"operating plan" at this point in our history.
15. Subcommittee Status Reports
===========================
PVL Subcommittee Status Report
Chairman: Mike Peterson
Mike presented a report on work in hand, and a number of
decisions that had been made during the meetings of this week
[include Mikes notes here when he sends them]
MVR Subcommittee Status Report
Chairman: Craig Humes. Report by John Merrill.
Event handling is an issue. Need to decide a model wide
mechanism or mechanisms.
Question still under consideration by the MVR subcommittee is:
How do you treat memory buffers and networks?
Bruce Haddon asked why this question of memory buffers and
networks is still a question.
John went on say that "mover-to-mover" protocols have yet to be
considered, together with the question of "negotiation". Dave
Skinner commented that this was important for being able to have
"inter-operable" movers.
Device specific issues were discussed with Rich Wrenn, including
formatting and other initialization. As well, the structure of
"cartridges" was revisited, leading to the thought that the
structure is the triple {cartridge, side, partition}, which is
what the PVL calls a "volume". The PVR mounts a {cartridge,
side} and the MVR loads a {partition}. From the PVL point of
view, the result is that it selects the complete triple.
There was concern that the PVL would select the "wrong" MVR for
a particular device, so there should be a mechanism for the
client to specify which mover it wishes to have used. This issue
was discussed for some time, with the intent that the issue
should be resolved in later PVL/MVR considerations.
PVL Subcommittee Report
Chairman: David Skinner
Dave showed a slide reporting that the question:
"Does the P1244.PVR support non-co-operating clients"
has been asked again, with the answer "yes".
Qualification: (1) P1244.PVR can be configured in such a
way that all clients must be co-operating.
(2) P1244.PVR can be configured in such a way that the
non-co-operating clients cannot interfere with
one another (i.e. partitioning). (3) P1244.PVR can be
configured in such a way that non-co-operating clients
can share resources (i.e., drives and cartridges) but no
protection against deadlock is provided.
Rationale: Eventually all storage systems may be P1244
compliant, today's systems are not and must be allowed
to evolve.
This question was then discussed at length. Bruce Haddon was of
the opinion that "sole access by PVL" is an overstatement of the
consistency requirements of the PVL.
# Action Item: the PVL/PVR committee will resolve this question.
This resolution is due by the end of the next meeting.
Dave Skinner reported that he will take up the technical editing
of the PVR (having obtained some agreement from his management).
16. Tabled Items
============
* Dates for 1995
Mar 14-17: DEC, Colorado Springs. (Host: Rich Wrenn) (done).
May 16-19: Valley Forge, Pennsylvania. (Host: Rich Garrison)
July 18-21: DEC, Seattle. (Host: Michael Peterson)
Sept. 14-15 (Monterey, at the end of the MSS)
Nov. 7-10 (tentatively ACSC, Los Angeles, Anit Gafni)
* Dates for 1996 (partial list)
Jan (towards the end of the month)
June (MSS is in Annecy, France).
17. Review of Open Action Items
===========================
1. Locate a replacement for swedishchef
Responsible: Dave Skinner
Due: 95/5 meeting
2. First draft of a process for "internal" balloting within the
SSSWG.
Responsible: Jack Cole
Due: 95/5 meeting
3. First draft of a process to address questions of lifetime, size,
logging, evaluation, etc. for RFP's.
Responsible: Dave Skinner and Jack Cole
Due: 95/5 meeting
4. RFP: "Plug and Play"
Responsible: David Skinner (PVR)
Due: 3/31
5. RFP: Naming conventions
Responsible: Craig Humes (MVR)
Due: 3/31
6. RFP: API calling conventions (argument passing)
Responsible: David Skinner (PVR)
Due: 3/31
7. RFP: Exception handling / Error propagation
Responsible: Craig Humes (MVR)
Due: 3/31
8. Status of the "draft" of the Reference Model (must it be
renewed, or is it a "guide" and does not need to be renewed).
Responsible: Merritt Jones.
Due: 95/5 meeting.
9. Copy of the operating plan outline and the subcommittee
operating plan to be posted to the reflector.
Responsible: Dave Skinner.
Due: 3/31
18. Proposed agenda for 95/5 meeting
================================
General Session
Announcements
Minutes of the previous meeting
Treasurer's report
Review of "Action Items from previous meeting
Status of recruitment and sponsorship.
Status of RFP's.
Report, discussion and decision of "Requirements for
Consistency" RFP (Mike Peterson)
Report and discussion of "Base Types" RFP (Mike Peterson)
Presentation and discussion of overview of VSS (Bob Baird)
Subcommittee Sessions
Closing General Session
Subcommittee status (each Chairman)
Tabled Items
Review of Open Action Items
Agenda for 95/7 meeting.
## Resolved: that this proposed agenda be accepted.
Moved: Jack Cole; 2nd: Merritt Jones
Carried unanimously.
19. Other business
============
* Action regarding POSIX PAR's
No change has occurred in POSIX activity since the last meeting.
No action is recommended.
* Process for internal balloting
An action item for issuing a draft to the reflector for
discussion has been agreed.
* Discussion of the DCE environment and standards
Joel Williams asked for clarification of the intent of this
item. The problem is that "DCE is not a referencable
* Recognition
Of Bruce Haddon, for again acting a recording secretary.
Joel Williams, for his "prime number" birthday.
Rich Wrenn, for hosting this meeting.
* comp.arch.storage
Mike Peterson asked whether we should recognize this USENET
group. Dave Skinner suggested we post our minutes and other
materials to this news group, as well as, perhaps other "comp"
groups. Merritt pointed out as an IEEE group, we are public, and
we should take advantage of appropriate channels of publicity.
20. Close of Meeting
================
The meeting was adjourned at 11:30 am, Friday, 17th March, 1995.
SSSWG Home